Hi Product Hunt! 👋 I’m the maker of Viaduct, a collaborative and AI-native architecture workspace built around the C4 model. I created it because integration designs often end up split between whiteboards, sequence diagrams, API specifications and documentation. These artifacts quickly drift apart, and static diagrams make it difficult to verify what actually happens between services. Viaduct keeps systems, containers, components, API contracts, message channels and documentation in one structured model. Its key feature is Magic Flow. It turns an integration scenario into an interactive, step-by-step flow. Each step references real architecture elements and can include HTTP endpoints, Kafka topics and model connections. Flows support alternative and parallel branches and can generate PlantUML sequence diagrams. Viaduct also provides an MCP server, allowing AI coding agents to work directly with the architecture model. Agents can explore systems and dependencies, read documentation and contracts, analyze Magic Flows, create or update architecture elements, and keep implementation decisions connected to the system design. Instead of giving an AI agent a static diagram or an outdated document, Viaduct provides structured architectural context it can query and update. While testing it on a banking transfer scenario, the player exposed a missing hop between Kafka and the notification service — exactly the kind of issue that is easy to overlook on a static diagram. I’d love feedback from software architects, system analysts and AI-assisted engineering teams: • How do you provide architectural context to coding agents today? • What would prevent you from replacing your current whiteboard workflow? • Which import format or MCP capability should Viaduct support next? Thanks for checking it out!

Viaduct

Executable architecture for humans and AI agents

Executable architecture for humans and AI agents